Forwarding and Control Element Separation (forces) Monday, November 18 at 0900-1130 ================================== CHAIRS: Patrick Droz David Putzolu Scribe: George Jones Agenda bashing: nothing changed Completed Last Calls draft-ietf-forces-framework-03.txt draft-ietf-forces-requirements-07.txt draft-ietf-forces-netlink-03.txt Discussion of draft-ietf-yang-model-01 - authors, history presented - motivation FE == Forwarding Element CE == Control Element * FE tells CE capabilities * FE tells CE current config * CE tells FE desired state - what is in the model * FE block (abstract base class) * Block library (Forwarding, QoS, filters, etc.) * Example FE Blocks * FE stage and directed graph * Two approaches in graph modeling * Topological (DiffServ) + No info carried forward * Topological (DiffServ) vs. Encoded State (QDDIM model) + Explicit info (preamble) carried forward to subsequent stages - open issues * Data modeling language: representation + SMI/SPPI/ASN.1/XML/UML ? * Topological vs. Encoded State approach * Modeling of actual functions + identify minimal categories/set of functions + model for each one - next steps * WG document * Data modeling language * define small set of functions Q: Is your intent to specify the way that the FE and CE communicate ? If so, it's very important to get right. A: yes. Q: (statement) Topological model is easier to reason about. A: We can take it on a case-by case basis. Q: (statement) We need to be aware that we don't just want to encode the a snapshot of the current state of the world A: We need help from the chairs to avoid this. Chair: show of hands, who's read ? Answer: not enough. Need more discussion to accept as a WG model. Abstract, not easy to discuss. Chair: We'd like people to start submitting actual protocol drafts. Easier to actually talk about/discuss. Action: model will not be WG document yet. Status of other drafts Closing: Do protocol submissions. End of meeting.